Questionnaire Processing with Artificial Intelligence

Questionnaire data is usually processed using statistical methods. How sophisticated meth-ods can be used is typically determined by the number of available responses; most meth-ods require at least 100 valid cases (completed questionnaires), some particularly strong methods require up to 1,000 cases. Statistical methods are usually considered robust and their results conclusive. However, statistical methods typically rely on the assumption of the linear independence of variables (which questionnaires try to follow) and the statistical methods seek, and often find, linear relationships between the variables.

Using machine learning (ML) algorithms of artificial intelligence (AI) works on the basis of different principles and produced different results. I want to emphasise, that AI processing of questionnaires should not be used as a substitute but as a supplement to statistical analysis. As it will perhaps become clearer from the below explanation of how the particular ML algorithm works, the relationships that AI can find, are not robust, in the statistical sense. The greatest benefits stem from using statistics and AI together. Statistical analysis cannot identify nonlinear patters, but if we identify these using AI, we can check their robustness using the traditional statistical methods.